Demo grpc service for trying out Mangle deductive database.

Mangle is an extension of Datalog. The repo is here: https://github.com/google/mangle

Building

go get ./...
go build ./...

Run the Server

This starts the gprc server, with a tiny database.

go run ./server --source=example/demo.mg

Run the Client

This queries the gprc server, with a recursive query.

go run ./client
go run ./client --query="reachable(X, /d)"

Alternatively: use grpcurl

grpcurl -plaintext -use-reflection=false -proto proto/mangle.proto \
  -d '{"query": "edge(/a, X)" }' localhost:8080 mangle.Mangle.Query

grpcurl can be obtained like so:

go install github.com/fullstorydev/grpcurl/cmd/grpcurl@latest

Regenerate the proto files

cd proto
protoc --go_out=. --go_opt=paths=source_relative \
  --go-grpc_out=. --go-grpc_opt=paths=source_relative mangle.proto

Persistence demo

Reminder: do not use this as a serious DB. The following is a demo deductive databse.

Create an empty db file

echo "0" | gzip - > /tmp/foo.mangle.db.gz

Start services with DB path

The following will load your (empty) DB and evaluate your program. Our demo program is the same, it contains only definitions of edges.

go run server/main.go --db=/tmp/foo.mangle.db.gz --source=example/demo.mg --persist=true

Add more edges

The client code only does querying, so we use grpcurl to send updates.

grpcurl -plaintext -use-reflection=false -proto proto/mangle.proto \
  -d '{"program": "edge(/d, /e). edge(/e, /f)." }' localhost:8080 mangle.Mangle.Update

When we run the query again, we find that our service now returns additional reachable nodes.

For every query, the service implementation does the computation of reachable nodes. If we know these queries in advance, we could also send an update that performs the computation.